Fraud Prevention Doesn’t Have an Awareness Problem Anymore — It’s Become an Infrastructure Problem
- João Moura, CEO and Co-Founder at Fraudio
- 26.06.2026 10:00 am #FraudPrevention #FinancialInfrastructure
It is difficult to find a bank, payment service provider (PSP), acquirer or fintech today that is not aware of the fraud problem. Nobody in the industry needs to be convinced that fraud is rising, that scams are becoming more sophisticated, or that criminals are using automation and AI to move faster than many traditional controls can respond. The problem really isn’t awareness, it’s fit.
Much of the fraud infrastructure still used across the payments ecosystem was designed for a different environment. It was built at a time where risk could be assessed inside relatively clear boundaries and where fraud could be identified through isolated transactions whilst controls could sit around specific parts of the payment journey, and now that world no longer exists.
Fraud today is networked and industrialised. Criminals don’t think in the same categories that businesses use to organise their risk teams. They don’t separate issuer fraud from acquirer risk, onboarding fraud from transaction monitoring, merchant abuse from AML, or account takeover from payment behaviour. They move across those boundaries precisely because the industry often still manages them separately.
Fraud has Industrialised, but Controls are Still Fragmented
For a long time, fraud prevention was often just seen as catching bad transactions. That view was understandable when many fraud patterns were simpler and slower. It was usually a stolen card used in an unusual location or a suspicious transaction amount.
The challenge now is that fraud has become a lot more operationally mature and refined. Criminal groups can now act and behave like businesses. They can test systems, learn from failed attempts, automate attacks and share tactics when one route becomes harder. AI is of course accelerating that trend by increasing the number of attacks and by improving their quality.
This changes the nature of the problem. The threat is now not about there being more bad transactions to stop, but that it is being organised through patterns that may only become visible when data is connected across accounts, cards, merchants, devices, behaviours and linked entities.
A transaction can look acceptable in isolation and still be part of a wider fraudulent pattern, and a merchant can appear healthy at onboarding and still begin to display risky behaviour later. Even a customer can pass a point-in-time check and still become compromised. The signal is often not in one event, but in the relationship between events and that is where fragmented systems start to fail.
The Stack Has Not Kept Up Speed With the Threat
The industry has made big steps of progress in fraud prevention, but many organisations are still limited by architectures that were not designed for modern fraud. They may have one system for transaction fraud, another for merchant monitoring, another for AML, another for disputes and another for onboarding. Each will be doing its job within its own boundary, but the fraudster is operating across the gaps in between them.
This creates a structural weakness. When controls are siloed, every part of the organisation only sees part of the risk. Teams make decisions based on incomplete context and models are trained on narrower datasets than they need. It can cause good customers to be blocked unnecessarily, while organised fraud can slip through because no single system has enough visibility to identify the wider pattern.
Essentially, the market has a fit problem. It’s not that firms don’t care about fraud or fail to invest in it. The issue is that adding more tools to a fragmented environment doesn’t necessarily create better intelligence and can actually in some cases just create more noise.
AI is Only as Strong as the Intelligence Behind It
There is a temptation to frame the current challenge as AI fraud vs AI fraud detection. That makes for an easy headline, but it misses the more important point.
AI is becoming a powerful tool for criminals because it helps them scale deception and adapt faster. However the answer is not simply for every business to add more AI. Poorly applied AI, trained on limited data or deployed inside a narrow workflow, will not solve the problem. It may only automate the limitations that already exist.
In payments, intelligence comes from context. It comes from understanding how entities behave over time, how they relate to each other, how payment patterns evolve, how merchants change, how accounts are linked and how risk signals move across the ecosystem. AI can be extremely powerful in this environment, but only when it is supported by the right data architecture, feedback loops and operational design.
Fraud Prevention is Becoming Growth Infrastructure
People sometimes see fraud prevention just as a loss reduction function, but nowadays that’s too limited for the way payments businesses operate. Fraud controls now affect onboarding, approval rates, customer experience, merchant portfolio quality, compliance workload and revenue growth.
A weak fraud system stops good activity from happening. It creates friction for legitimate customers, delays merchant onboarding, generates unnecessary manual reviews and causes businesses to become more conservative than they need to be. When risk teams lack confidence, growth slows down.
This is especially important for acquirers, PSPs and fintechs operating in competitive markets. They need to approve more good transactions, onboard strong merchants faster and expand into new segments without exposing themselves to unacceptable levels of fraud, abuse or compliance risk. That is not possible if fraud prevention is treated as a defensive layer at the edge of the business.
The Future is Connected, Continuous and Ecosystem-Aware
If fraud is networked, the response has to be networked too. That does not mean every firm must expose all of its data to everyone else, or does it mean creating uncontrolled shared databases without proper governance. It means recognising that fraud prevention improves when intelligence is broader, more structured and more connected.
The next model has to bring together fraud detection, merchant monitoring, AML signals, behavioural analysis and operational feedback into a more coherent view of risk. It needs to understand whether a single payment looks suspicious and whether the wider pattern makes sense.
Fraud prevention isn’t just a security function anymore, it’s now core to payments infrastructure. The sooner the industry treats it that way, the better prepared it will be for the next generation of financial crime.
